Abstract:

Assessing the quality of experience of video services takes into account the subjective feelings of viewers. Estimating such a rating using machine learning models can be important from the operator's point of view, as the feedback can be used to improve the level of service provided. Customer feedback also takes into account aspects of the service that are not visible in traditional monitoring systems. In order to train supervised machine learning models, labels are needed, which are best collected from the customers of the service, as it is the subjective evaluation that concerns the customer's perception of the service. This short paper describes an approach to collect ratings from customers of an adaptive streaming based TV service of the Polish telecom operator INEA. These ratings are then used to train a machine learning model that correlates the subjective ratings with data from a central logging system.